generic advertising without supply control: implications of funding

Generic advertising without supply control:implications of funding mechanisms for

advertising intensities in competitive industries{

John W. Freebairn and Julian M. Alston*

Producer pro¢t-maximising rules for generic commodity advertising programs andassociated funding levies are derived. Lump-sum, per unit and ad valorem levies,and government subsidy funding arrangements are compared and contrasted. Theinitial single-product competitive market model is extended to incorporate inter-national trade, government price policies, and multiple commodity interactions.

1. Introduction

Generic commodity advertising programs are an important feature ofAustralian agriculture (IAC 1976) and of agriculture in other countries (inthe United States, for example, Forker and Ward (1993), and Vande Kampand Kaiser (1999), estimate annual expenditure at over US$1 billion). Manycommodities have some generic advertising (especially livestock products,including dairy, meats, wool and eggs, and tree crops such as apples, citrus,and various nuts). For some commodities, producer levies to fund genericadvertising have been as high as 2 per cent of gross sales, and in some cases,especially where exports are involved, governments have provided matchinggrants. This article reviews and extends models for determining genericadvertising expenditure, and associated producer levy rates, that wouldmaximise net returns or pro¢ts to commodity producers, and clari¢es thee¡ects of di¡erences in funding mechanisms and market situations on theproducer optimum.1

# Australian Agricultural and Resource Economics Society Inc. and Blackwell Publishers Ltd 2001,108 Cowley Road, Oxford OX4 1JF, UK or 350 Main Street, Malden, MA 02148, USA.

The Australian Journal of Agricultural and Resource Economics, 45:1, pp. 117^145

{We are grateful for the useful comments provided by anonymous referees, and JenniferS. James on an earlier version of the article.

* John W. Freebairn, Department of Economics, the University of Melbourne, Australia.Julian M. Alston, Department of Agricultural and Resource Economics, University ofCalifornia, Davis, USA.

1No assessment is made about the more controversial implications of advertising for thewelfare of buyers and of society.

There is a very large literature on modelling the choice of pro¢t-maximisingadvertising. Most of the industrial organisation literature in this areaconsiders the case of ¢rms with market power, including monopoly, oligopolyand monopolistic competition, where ¢rms have control over supply as wellas advertising. In the context of generic advertising of agricultural com-modities, our interest is with ¢rms acting as price takers, and the advertisingchoice takes a competitive supply response as beyond the control of theorganisation in charge of the advertising and other promotional decisions.Models of generic advertising also vary in terms of their treatment of the

source of funding. Here we distinguish between whether the funds areregarded as a lump-sum charge, as was initially studied by Nerlove andWaugh (1961) and more recently by Goddard and McCutcheon (1993), orcollected as a levy. We draw a further distinction between a speci¢c (per unit)levy and a percentage (ad valorem) levy. IAC (1976), De Boer (1997) andChang and Kinnucan (1991) present diagrams to represent a model of levy-based promotion, and Alston et al. (1994), Alston et al. (1998), andKinnucan (1999a, b) provide more formal analyses. In addition, we evaluatethe case where the government provides a grant to supplement the producerlevy (see also Kinnucan and Christian 1997). Not surprisingly, the source offunds a¡ects the pro¢t-maximising level of generic advertising and levyrates.Most of the reported models consider the case of a closed economy or

non-traded commodity. Kinnucan (1999a) and Cran¢eld and Goddard(1999) speci¢cally allow for trade, and this is a more realistic framework forconsideration of generic advertising for most agricultural commodities inmost countries, including Australia. As we show, recognition of the exportor import status of an agricultural commodity in£uences the producer pro¢t-maximising level of generic advertisingA number of other model extensions to capture greater realism have been

considered. These include explicit allowance for the e¡ects of policy inter-ventions, for example, Goddard and Tielu (1988) and Kinnucan (1999b), andmulti-stage production systems, for example, Wohlgenant (1993). Of recentinterest, and with likely important implications, are multi-product modelsthat recognise a number of cross-product e¡ects of generic advertising£owing from one product to another. Here the studies by Piggott et al.(1995), Hill et al. (1996), Kinnucan (1996) and Kinnucan and Miao (2000)point to the di¡erent cross-product price and advertising e¡ects in£uencingproducer returns, but they do not formally derive pro¢t-maximising levels ofgeneric advertising and associated levy rates.In this article we synthesise these various elements within a uni¢ed frame-

work to show the di¡erences in results for producer's optimal advertisingexpenditures, with di¡erent funding mechanisms, and under a range of

118 J.W. Freebairn and J.M. Alston

# Australian Agricultural and Resource Economics Society Inc. and Blackwell Publishers Ltd 2001

market and policy situations. The rest of the article is organised as follows.Diagrams and algebra are used to derive pro¢t-maximising advertisingstrategies. Initially, we consider a single commodity, closed economy, withno government intervention. For this case we derive producer pro¢t-maximising advertising and levy rates for the cases of lump-sum funding,and funding by per unit and ad valorem levies; and, in the case of a per unitlevy, we allow for a matching government subsidy. The next section addsan international dimension, with most of the emphasis on an exportcommodity. Then we compare and contrast the marginal conditions andelasticity rules for the producer pro¢t-maximising advertising levels and levyrates derived under the di¡erent funding options and trade status. Weillustrate how government policy interventions can be incorporated into themodels and how they can alter producer pro¢t-maximising choices. Then wenote some of the likely implications of taking into account multiple-productinteractions as they a¡ect producer pro¢t-maximising generic commodityadvertising strategies, and suggest further directions of analysis. Finally wediscuss some implications of our analysis for the speci¢cation of econometricmodels of the demand response to advertising and summarise the main¢ndings.

2. Profit-maximising advertising in a closed economy

We begin with a single-market model of supply, demand, and marketequilibrium for a commodity, say, wool or bananas, and then we specifyspecial features for the di¡erent options to fund advertising expenditure.Functions for demand, supply and market clearance speci¢ed at the farmlevel are given by:2

Qd � f �P;A; Zd� �1�Qs � g�Pp; Zs� �2�Q � Qd � Qs �3�

where Q is quantity supplied and demanded, P is buyer price, Pp is producerprice, A is the advertising expenditure,3 and Zd and Zs are demand andsupply curve shift variables (which are ignored in the remainder of the

2 The demand function at the farm level can be considered a derived demand taking intoaccount retail demand and the farm-to-retail marketing activities of transport, storage,processing, and distribution.

3 In this article we take A to measure the opportunity cost of funds. This means we candrop the r term found initially in Nerlove and Waugh (1961) and some subsequent work.

Generic advertising without supply control 119


article). Advertising, if e¡ective, shifts out the demand curve, that is@Qd=@A > 0, and it may reduce (or increase) the elasticity of demand, that is@2Qd=@P@A > �or <� 0.The way in which the advertising is funded in£uences the producer

pro¢t-maximising expenditure on advertising, because it in£uences the shareof advertising costs ultimately borne by the producers. Under a lump-sumfunding arrangement, producers both pay and bear all the ¢nal economicincidence of the costs of advertising, and the net producer price is the marketprice �Pp � P�. In the case of a per unit (¢xed, or speci¢c) levy, for example,$T per kilogram of wool or bananas, advertising expenditure is equal to thelevy times output �A � T Q�, but some of the economic costs are passed onto consumers, and the net producer price is the market price minus the levy�Pp � Pÿ T �. With an ad valorem levy at a rate t, for example, t � 100 percent of the value of wool or banana sales, advertising expenditure is equal tothe levy times the value of sales �A � tPQ�, and the net producer price is afraction �1ÿ t� of the market price �Pp � �1ÿ t�P�. Finally, when thegovernment provides a subsidy, for example, $x per $ of producer fundsraised by a levy collected from growers, the advertising expenditure is greaterthan both the levy revenue and the producer cost.In all the cases below, both the advertising expenditure and the associated

levy rate, if relevant, are chosen to maximise producer pro¢t or quasi-rent.Pro¢t, p, is

p � PQÿ T V Cÿ A �4�where P and Q are price and quantity determined by market equilibrium ofequations (1), (2) and (3), A is advertising expenditure from producer fundsraised by commodity levies or lump-sum assessments, and T V C is totalvariable cost, or the integral of the competitive supply curve in equation (2)so that:

T V C �Z

gÿ1�Pp�dQ: �5�

(Implicitly we assume the absence of avoidable ¢xed costs or no change inavoidable ¢xed costs when we measure changes in pro¢t as changes inproducer surplus.) In the model solutions below we use the converse resultthat dT V C=dQ �MC (where MC is marginal cost), and under competitivebehaviour producers choose output to equate MC and Pp, so that MC

corresponds to the supply curve, given by equation (2). With this modelframework, we consider the pro¢t-maximising advertising expenditure, A,and levy rates T and t if relevant, in turn for each of the advertising fundingoptions.



2.1 Lump-sum funding

Nerlove and Waugh (1961) ¢rst developed the optimal rule for advertisinggiven lump-sum funding. Figure 1 illustrates the story. Initially, D0 isdemand, S0 is supply, P0 and Q0 are the market-clearing price and quantity,and area aP0e is producer surplus or industry pro¢t.An increase in e¡ective advertising (by dA) shifts the demand curve out.

One perspective shown in ¢gure 1 is a parallel outwards shift to D1, withef � �@Q=@A�dA. The new equilibrium is at price P1 and quantity Q1, andgross producer surplus increases by area P0P1ge. Alternatively, advertisingalso may make demand less elastic, say, shifting the demand curve from D0

to D2 (passing through point f ). Note that, in comparing the shift of demandto D2 rather than D1, a case where demand becomes more inelastic as wellas the curve shifting outwards, the new equilibrium at h is at a higher priceand quantity, with a larger increase in gross producer surplus. If the supplycurve at some time in the future were to shift downwards, say, because ofR&D, a more inelastic demand might mean lower producer surplus in time,as was pointed out by Quilkey (1986). For completeness, we could considerother advertising strategies that shift demand out, but not as far as point f ,and make it less elastic, with the new price-quantity equilibrium outcome

Figure 1 Lump sum funding model



being to the left or right of point g; or advertising might result in a moreelastic demand curve.Intuitively, the pro¢t-maximising advertising expenditure would be that

amount where the marginal gross bene¢t from additional advertising, givenby �dP=dA�Q, equals the marginal cost of advertising, which is one�dA=dA � 1�. An expression for the pro¢t-maximising lump-sum expenditureon advertising can be derived more formally by solving for the value of A

that will maximise pro¢t p in equation (4) subject to equations (1), (2) and(3). Taking the derivative of p in equation (4) and setting it to zero:4

dpdA� 0 � P

dQ

dA� Q

dP

dAÿ dT V C

dAÿ 1 �6�

Now, expanding �dT V C=dA� � �dT V C=dQ��dQ=dA�, recognising thatdT V C=dQ �MC � P, the ¢rst and third right-hand terms of equation (6)cancel, and the ¢rst-order necessary condition from equation (6) can beexpressed as:

QdP

dA� 1: �7�

In equation (7), the left-hand side represents the marginal revenue gain frommore advertising, and the right-hand side is the marginal cost of a dollarincrease in the lump-sum advertising budget.An expression for dP=dA in equation (7) can be obtained by taking total

derivatives of the market demand and supply curves (1) and (2) and equatingthem via equation (3).5 After substituting for dP=dA in equation (7), andsome rearrangement, we obtain:

Q@f

@A� @g@Pÿ @f@P

�8�

which can be used to ¢nd the optimal value for A, given the marginal e¡ectof advertising on demand, @f =@A, and the price slopes of demand and supply,@f =@P and @g=@P. Alternatively, equation (8) can be re-expressed as:

fL S � A

PQ� a

e� Z�9�

4 For simplicity we assume an interior solution. As noted by a referee, in some cases theoptimum solution will be a corner solution of zero advertising. An implicit assumption whenwe say that we have found a maximum is that the marginal return from the advertising isdownward sloping (i.e., that @2Q=@A2 < 0) which means that the advertising elasticity ispositive but less than one (i.e., 0 < �@Q=@A��A=Q� < 1).

5 Full details of the derivation of these and other formulae are available from theauthors.



where a � �@f =@A��A=Q� is the elasticity of demand with respect toadvertising, Z � ÿ�@f =@P��P=Q� is the absolute value of the own-priceelasticity of demand (i.e., Z > 0), e � �@g=@P��P=Q� is the elasticity of supply,and fL S � A=�PQ� is the pro¢t-maximising advertising intensity (advertisingexpenditure as a share of total revenue), with all elasticities measured at thepro¢t-maximising equilibrium point. Equation (9) is the Nerlove^Waugh rulefor choosing a pro¢t-maximising, lump-sum advertising budget.

2.2 Per unit levy funding

In most cases generic advertising of agricultural products is funded by a levyon output collected from producers (or the ¢rst handler of a farm product),either a ¢xed (or per unit quantity) levy, or an ad valorem levy. Figure 2illustrates the situation for a ¢xed levy. As for ¢gure 1, in ¢gure 2, in theinitial situation, D0 is demand, S0 is supply, P0 and Q0 are the market-clearing price and quantity, and area aP0e is producer surplus or industrypro¢t.Collection of a levy, which is used to fund generic advertising, can be

analysed as a shift of both the demand curve and the supply curve.Advertising, as before, shifts demand outwards to D1. At the same time, the

Figure 2 Per unit levy funding



per unit levy adds to variable and marginal costs, with the latter movingthe supply curve upwards by the amount of the levy to S1. A new equilibriumis established with a higher price and quantity, at P1 and Q1, respectively.Producer surplus becomes area bP1g�� akm�. The net gain in producersurplus is bP1gÿ aP0e, or area P0mke. The levy will increase so long as theadditional sales revenue gained exceeds the extra levy costs and the increasedcosts of production associated with greater output. At the optimum, thevertical increase in demand (from a marginal increase in advertising) justbalances the vertical shift in supply (from an increase in the levy to ¢nancethe marginal increase in advertising) so that the quantity produced andconsumed, and producer surplus, do not change (i.e., dP=dA � dP=dT anddQ=dT � 0).We wish to derive an expression for the levy, T (and the associated

advertising expenditure, A, that will maximise p in equation (4) givenequations (1), (2), and (3), and subject to the constraints that advertisingexpenditure is equal to the amount of revenue raised by the levy (i.e.,A � T Q) and that net producer price is Pp � Pÿ T . To do this, wesubstitute for A � T Q in equation (4), take the derivative of p with respectto T , and set it to zero, which gives the pro¢t-maximising ¢rst-ordernecessary condition:

dpdT� 0 � P

dQ

dT� Q

dP

dTÿ dT V C

dTÿ Qÿ T

dQ

dT: �10�

Then, using dT V C=dT � �dT V C=dQ� �dQ=dT �, and recognising thatdT V C=dQ � Pp � Pÿ T , the ¢rst, third, and last terms on the right-handside cancel, and the ¢rst-order condition in equation (10) can be expressedas:

QdP

dT� Q; or

dP

dT� 1: �11�

In equation (11), the left-hand side �QdP=dT � is the marginal gain in revenuefrom extra advertising funded by an incremental increase in the levy, andthe right-hand side �Q� is the marginal cost of an increase in the levy. At themaximum, a marginal increase in the levy to fund advertising increasesmarginal cost by the same amount as the extra advertising in shifting outdemand increases price, and hence dQ=dT � 0, and producer surplus (orpro¢t) is unchanged.To obtain an expression for dP=dT in equation (11), we take the total

derivatives of the market demand and supply curves (1) and (2), recognisingthat PP � Pÿ T and A � T Q, and equate them using the market clearingidentity (3). After substituting the resulting expression for dP=dT in (11),and cancelling some terms, we obtain:



Q@f

@A� ÿ @f

@P�12�

Further manipulation of equation (12) to obtain elasticities leads to thefollowing condition for the pro¢t-maximising levy and advertising intensity:

fPU � A

PQ� T

P� a

Z�13�

where fPU is the optimal advertising intensity funded by a per unit levy,T =P is the levy as a proportion of the market price, and, as before, a is theadvertising elasticity and Z is the absolute value of the demand priceelasticity.Note that, coincidentally, equation (13) is the same as the Dorfman^

Steiner advertising rule, but it is derived for very di¡erent circumstances andthe parameters represent di¡erent forces. Dorfman and Steiner (1954)considered the case of a price-maker who chooses production and advertisingto equate marginal cost and marginal revenue, with marginal revenue belowmarket price. Essentially, assuming a constant price and marginal cost (or,less strongly, a constant margin of price less marginal cost) around thepro¢t-maximising point, the seller's bene¢t from extra sales driven byadvertising (measured by the advertising elasticity), is equal to price lessmarginal cost. For a monopolist, the margin of price minus marginal cost isequal to price times the inverse of the demand elasticity. In contrast, thecontext here is one of collective action in advertising by a set of price-takingproducers without supply control powers, such that production is chosen toequate marginal cost with price, not the industry marginal revenue. At thepro¢t-maximising advertising level and levy rate, quantity is constant, thatis, dQ=dT � 0. The inverse of the demand elasticity is used to convert thequantity-expansion e¡ect of advertising into a price-increase e¡ect, and it isthe price-increase e¡ect that determines the bene¢ts from generic advertisingfor the industry with competitive producers.

2.3 Ad valorem levy funding

Compared with the per-unit levy, which causes a parallel shift upwards ofthe supply curve, an ad valorem levy results in a proportional shift of thesupply curve, but otherwise the story is similar. We wish to derive anexpression for the levy, t, that will maximise p in equation (4) givenequations (1), (2), (3), and subject to the constraints that advertisingexpenditure is equal to the amount of revenue raised by the levy (i.e.,A � tPQ) and that the producer price is given by Pp � P�1ÿ t�. To do this,we substitute for A in equation (4) and take the derivative of p with respect



to t, and set it to zero, which gives the ¢rst-order necessary condition forpro¢t maximisation:

dpdt� 0 � Q�1ÿ t� dP

dt� P�1ÿ t� dQ

dtÿ dT V C

dtÿ PQ: �14�

Expanding the second last right-hand term ö using dT V C=dt ��dT V C=dQ��dQ=dt� � P�1ÿ t��dQ=dt� ö the ¢rst-order condition fromequation (14) can be expressed as:

dP

dt� P

1ÿ t: �15�

To obtain an expression for dP=dt in equation (15) involves taking totalderivatives of the demand and supply equations (1) and (2), and usingA � tPQ and Pp � �1ÿ t�P. Substituting for dP=dt in equation (15), the ¢rst-order condition can be restated as:

@f

@AQ � ÿ @f

@f�16�

which can be expressed in terms of the pro¢t-maximising advertisingintensity and elasticities of demand with respect to price and advertising as:

fAV � A

PQ� t � a

Z�17�

where fAV is the optimal advertising intensity funded by an ad valorem levy, t,and all the terms are as de¢ned above. Note that the producers' optimal levy isthe same, regardless of whether it is speci¢ed as per unit or ad valorem, whichcan be seen by noting that t � T =P and comparing equations (13) and (17).

2.4 Government subsidy for promotion

In some cases governments top up or provide a matching grant forindustry-provided funds for advertising, e¡ectively an advertising subsidy.For example, in Australia prior to 1993^94, producer wool promotion levieswere matched dollar for dollar, and in Japan governments contribute to thegeneric advertising of £uid milk (Suzuki et al. 1994). In the United States,subsidies have applied for export promotion under the Temporary ExportAssistance (TEA) program, established under the 1985 farm bill, and theMarket Promotion Program (MPP) established under the 1990 farm bill.6

6 Ackerman and Henneberry (1992) discuss the programs, and various papers in Nicholset al. (1991), for example, provide more speci¢c analysis of US export promotion programswith support from the TEA program. Vande Kamp and Kaiser (1999) provide more recentdata.



This kind of government subsidy could be speci¢ed as a lump sum, or asa per unit or ad valorem rate. Here, we consider an alternative case wherethe government provides a matching grant in proportion to the amountraised by a per unit levy, raising the advertising budget from A � T Q toA � �1� x�T MQ where x is the proportion of the budget provided by thegovernment, and the superscript M denotes that this levy rate will di¡er fromthe one derived in the absence of the matching funds.7 In essence, the pro-ducer cost per dollar of advertising is reduced from one dollar to 1=�1� x�dollars; alternatively, advertising generates a larger demand expansion perdollar of producer levy.We follow the procedures for deriving the pro¢t-maximising levy T from

the section `Per unit levy funding' above. The pro¢t-maximising amount ofadvertising is given by similar ¢rst-order conditions as in the absence of thematching support ö i.e., dQ=dT M � 0, and dP=dT M � 1. The marginalcondition in terms of slope parameters becomes:

Q@f

@A�1� x� � ÿ @f

@P�18�

and the elasticity rule for the advertising intensity becomes:

fM � T M

P�1� x� � a

Z�1� x�: �19�

In terms of the optimal check-o¡ rate, the form of the rule is unchanged.That is:

T M

P� a

Z�20�

where, as before, T M is the levy per unit collected from producers, a is theelasticity of demand with respect to advertising, and Z is the absolute valueof the price elasticity of demand. However, now these elasticities are de¢nedfor the equilibrium with a matching grant, which means a higher advertisingintensity, fM � �1� x��a=Z�, given a higher advertising expenditure, AM ��1� x�T MQ, which is funded partly by the levy and partly by the governmentsubsidy, x.The levy rate T in equation (13) without the government subsidy (or with

x � 0) can be compared with the levy rate T M in equation (20) with a subsidy(or with x > 0). For a common demand elasticity (i.e., Z � ZM), the optimal

7Alternatively, the subsidy could be represented as T �X. Similarly, for an ad valoremindustry levy t, the subsidy could be represented as t�1� x� or as tP�X. Likewise,government subsidies could be used to augment producer lump-sum funding of genericadvertising.



levy rate �T M=P� implied by the matching grant is greater than (less than)the levy rate with no matching grant �T =P� if the advertising elasticity withthe matching grant, aM, is greater than (less than) the advertising elasticitywith no matching grant, a. With a larger total advertising budget, that is,AM � T M�1� x�Q > T Q � A, given diminishing marginal returns, themarginal return to advertising in the pro¢t-maximising area is lower, that is@2Q=@2A < 0, a result broadly found in the quantitative literature whenallowed (Lambin 1976, and Forker and Ward 1993) and a result required tomeet second-order conditions for pro¢t maximisation. Even though theadvertising intensity is higher �fM > f�, if the advertising elasticity, a, is adeclining function of the amount of advertising, then, aM < a. In turn, thismeans that the introduction of a matching government grant implies a lowerpro¢t-maximising producer levy rate, but a higher total advertisingexpenditure, and these e¡ects will be larger, the larger is the subsidy, x.Whether the government grant crowds out producer spending to the extentof reducing total levy revenue is less clear, since the lower rate of levy isapplied to a larger total value of sales revenue.

3. International trade

The single-market model might be regarded as a closed economy, or non-traded product model. In reality, most agricultural commodities are tradedinternationally. For export goods, the analysis must distinguish betweenadvertising that applies only to the domestic market, only to the exportmarket, or to both. And care must be given to the de¢nition of the tax base forcollecting levies to fund the advertising. Both the target market and the levybase will in£uence the answer. Also of interest is the comparison of pro¢t-maximising advertising and levy rates based on a closed-economy model whenthe commodity is actually traded. Kinnucan (1999a), Cran¢eld and Goddard(1999), and Kinnucan and Myrland (2000) have developed models where theadvertised commodities are exported or imported. To illustrate some keyimplications of explicit recognition of international commodity trade, in thissection we extend the model to encompass both export sales and domesticsales, and derive producer pro¢t-maximising rules for a per unit levy on totalproduction to fund advertising on the domestic market.Figure 3 illustrates the case for a homogeneous product sold domestically

and exported. For simplicity, free trade is assumed and transport costs areignored. Then, the same producer and buyer prices apply for domestic andexport sales. Before advertising, domestic demand is Dd

0, export demand isDe

0, and total market demand is D0 � Dd0 � De

0. With initial supply S0, marketequilibrium is given by price P0, quantities Qd

0, Qe0, and Q0 � Qd

0 � Qe0, and

initial producer surplus is area aP0e.



Figure 3 Per unit levy funding of domestic market advertising

Generic

advertisingwithout

supplycontrol

129

#Australian

Agriculturaland

Resource

Econom

icsSociety

Inc.andBlackw

ellPublishers

Ltd

2001

Now, suppose a levy, T , is collected on all sales, Q, to provide advertisingfunds, Ad � T Q, which are used to promote domestic sales. The domesticdemand shifts out to Dd

1, export demand does not change, and total demandshifts out to D1 � Dd

1 � De0. The levy shifts the supply curve upwards to

S1 � S0 � T . The new equilibrium price is P1, domestic sales rise to Qd1,

export sales fall to Qe1 and total sales rise to Q1. Producer surplus rises by

P0mke, as was the case in ¢gure 2 where, for convenience, the same notationhas been used for ¢gure 2 and the third panel of ¢gure 3, representing thetotal market.Focusing on the third panel of ¢gure 3, and analogous to the discussion

in section 2 and ¢gure 2, the levy T should be raised to fund advertising onthe domestic market to the point where the marginal bene¢t to producersfrom the advertising-induced increase in price �QdP=dAd� is equal to themarginal cost to producers of the levy �QdT �. This, in turn, gives the ¢rst-order condition for pro¢t maximisation:

dP

dT� 1: �21�

This is the same form as equation (11), except we now interpret P as themarket price that clears domestic and export sales. An analogous set ofarguments could be made for using the levy to fund export promotion, ormore generally both domestic and export promotion.A key result in both equation (11) for the non-traded good, and equation

(21) for the traded good, is that for advertising to be pro¢table in a com-petitive industry (with upwards-sloping supply), it must induce an increase inprice. Indeed, since a levy of T per unit increases average costs by T per unit,the resulting advertising expenditure must give rise to an increase in averagerevenue of at least T per unit. The result in equations (11) and (21) is anoptimum since, at the margin, the increase in producer revenue is exactlybalanced by the increase in producer costs but, for inframarginal quantitiesof advertising and levies, because of diminishing returns, the marginal bene¢texceeds the marginal cost.A formal expression for dP=dT for equation (21) can be derived from the

traded-good commodity model, de¢ned by:

Qd � fd�P;Ad�;where Ad � T Q �22�Qe � fe�P�; �23�

Q � g�Pp�;where Pp � Pÿ T �24�Qd � Qe � Q �25�

where equation (22) is the domestic demand equation with domesticadvertising equal to Ad, equation (23) is the export demand equation,



equation (24) is the supply equation as in (2), and equation (25) is the marketclearing identity. Taking total derivations of equations (22) to (24), and usingequation (25), an expression can be obtained for dP=dT .After substituting the expression for dP=dT into equation (21), and some

further manipulation, the optimal advertising intensity for an export good,fE, can be derived similarly to equation (13) as:

fE � Ad

PQ� T

P� wda

d

wdZd � �1ÿ wd�Ze

�26�

where wd denotes sales on the domestic market as a share of total sales(i.e., the quantity share wd � Qd=Q or, equivalently, the value sharewd � PQd=PQ�, ad � �dQd=dAd��Ad=Qd� is the elasticity of domestic demandresponse to advertising, and Zi � ÿ�dQi=dP��P=Qi� is the absolute value ofthe own-price elasticity of demand in market i (for i � d or e, representingthe domestic and export demands, respectively).In equation (26), the pro¢t-maximising advertising intensity and levy rate

will be greater the larger is the sales response to advertising (the larger is ad),the less elastic is demand in both the advertised and non-advertised segments(the smaller are Zd and Ze), and the more important is the advertised marketin total sales (the larger is wd). The same equations could be used to representexport market advertising funded by a levy on all of production, simply byreversing the roles of the two markets (i.e., switching the indexes, d and e inthe equations). Further, a model with both domestic and export advertisingcould be represented, approximately, by simply adding the two componentstogether.Further inspection of ¢gure 3 and equation (26) enables several obser-

vations to be drawn about pro¢t-maximising levels of advertising on eitherthe domestic or export markets. For the situation of a small-countryproducer (in the sense that its export demand curve is perfectly elastic, or inequation (26) Ze � 1), no amount of domestic advertising or exportadvertising is worthwhile, with absent government intervention in thecommodity market. This is because, in such a case, no matter how e¡ectiveadvertising is in increasing the domestic demand, it cannot give rise toincreases in the domestic and world price. Policy interventions and otherforces that separate the markets and break the law of one price can alter thisstory (as illustrated in ¢gure 4). However, in the absence of such inter-ventions, it is unlikely to be pro¢table to promote domestic sales ofagricultural products that are also sold on export markets where the countryfaces a highly elastic export demand. The same is true for goods thatcompete with highly elastically supplied imports. For instance, in theAustralian citrus industry, the promotion of fresh oranges on the domestic



market might give rise to an increase in demand and quantity sold. But sinceAustralia is a small country in the world market for orange juice, the maine¡ect is likely to be a diversion of Australian oranges from processing to thefresh market, with little if any impact on price or average revenue.8 Even whenprices are not exogenous, in a multimarket setting arbitrage e¡ects willdampen the price-enhancing e¡ects of advertising in one market segment,diminishing the producer returns and optimal advertising expenditure relativeto what would be implied by a single-market analysis as shown above.The results of this section can be used to choose levy rates to fund

advertising of domestic sales and export sales, or to allocate a givenadvertising sum between the two markets.9 Clearly, this type of analysis canbe extended in many ways. Examples include more than two market seg-ments, or markets may be segmented by product type ö such as the variousdairy products and fruit products ö as well as geographically.10 As before,situations of lump-sum funding and of matching subsidies also could beanalysed. And, as shown in Kinnucan (1999a), these procedures can beapplied to a commodity that is imported.

4. Some comparisons

At this point it is useful to compare and contrast the rules derived forproducer pro¢t-maximising levels of advertising, and associated levy rates,from the di¡erent models presented in the preceding two sections. Table 1provides a summary of ¢rst-order conditions that are used to derive thepro¢t-maximising levels of advertising, and of the advertising intensitiesexpressed as functions of elasticities at these optimum levels, for a number ofthe models. The models di¡er with respect to assumptions about trade status(i.e., non-traded versus export), and about funding options (i.e., lump sum,per unit levy, ad valorem levy, and per unit levy with a matching government

8 This fact did not prevent the citrus fruit marketing board from spending levy funds onfresh citrus promotion in the 1980s, although, interestingly, the advertising intensity washigher in the Riverina than in Melbourne.

9 In a typical case where the export demand is relatively elastic, raising the price on bothmarkets through levy-funded export promotion has some elements in common with pricediscrimination against the more inelastic domestic market. When the `promotion' meansprice discounts rather than advertising, then the arrangement is in e¡ect an export subsidy¢nanced by an output tax, and this is identical in e¡ect to a price discrimination and poolingarrangement (e.g., see Alston and Freebairn 1988, for discussion). The same may be true ifthe two markets are for di¡erent end-uses rather than di¡erent markets for the same end-use.

10 The conference volume edited by Nichols et al. (1991) is devoted to generic promotionprograms for agricultural exports. Also, see Goddard and Conboy (1993).



subsidy). There are some important similarities, and some subtle contrasts,in the derived pro¢t-maximising advertising levels and levy rates for thedi¡erent sets of assumption.For the typical agricultural commodity production system, with many

competitive producers and no industry-wide supply control, advertising canimprove producer returns only if it results in an increase in market price.Producers gain primarily from the higher price on existing sales, and to a lesserextent from the increases in output. A number of factors in turn in£uence thepotential producer price-enhancing e¡ect of advertising. First, the moree¡ective is advertising in increasing demand at any price, ceteris paribus, thelarger will be the price gain and the pro¢t-maximising advertising intensity. Inall cases, the optimal advertising intensity increases with the advertisingelasticity, a. Second, elasticities of demand and supply determine the price-increasing e¡ect of both advertising-induced increases in demand and levies tofund the advertising. In particular, when demand is less elastic, a given demand

Table 1 Profit-maximising advertising rules: effects of funding methods and trade status

Funding method ExpenditureFirst-orderconditions Advertising intensity and levy rate

Closed economy

Lump sum, A A QdP

dA� 1 fL S � A

PQ� a

e� Z

Per unit levy T T QdP

dT� 1;

dQ

dT� 0 fPU � A

PQ� T

P� a

Z

Ad valorem levy, t tPQdP

dt� P

1ÿ t;dQ

dt� 0 fAV � A

PQ� t � a

Z

Per unit levy, T M andMatching subsidy, x

�1� x�T MQdP

dT� 1;

dQ

dt� 0 fM � A

PQ� T M

P�1� x� � a

Z�1� x�

Export market

Lump sum, A A QdP

dA� 1 f � A

PQ� wda

d

e� wdZd � �1ÿ wd�Ze

Per unit levy, T T QdP

dT� 1 fE � A

PQ� wda

d

wdZd � �1ÿ wd�Ze

Source: Derived from text. P is price, Q is quantity, A is total advertising expenditure, T is the per unitlevy, t is the ad valorem levy, and x is the matching government subsidy. For the closed economy, a isthe elasticity of demand with respect to the advertising expenditure, Z is the absolute value of the priceelasticity of demand, e is the price elasticity of supply, and the M superscript distinguishes the matchinggovernment subsidy case. For the export market case with advertising of domestic sales, wd is the shareof the product sold domestically, ad is the advertising elasticity for domestic sales, and Zd and Ze arethe absolute values of the price elasticities of demand for domestic and export sales, respectively.



shift in the quantity direction translates into a greater shift in the price direction,and thus, everything else equal, the advertising is more pro¢table for producers.Hence, in table 1, for a nontraded commodity the advertising intensity isinversely related to the absolute value of the elasticity of demand, Z.For an export commodity, we can see that the result for the overall

advertising intensity is equivalent, given that the overall elasticity of demandis equal to the share-weighted sum of the domestic and export marketelasticities of demand, i.e., Z � wdZ

d � �1ÿ wd�Ze. Hence, the optimal levyrate decreases with increases in either the elasticity of demand in the marketwhere the advertising applies, or the elasticity of demand in the other marketfor the good. Further, if the export demand is relatively elastic (as is thetypical case), the optimal levy rate increases with an increase in the domesticmarket share (because this means a less elastic overall demand). In theextreme case of a small country �Ze � 1�, or even in a less-extreme case ofheavy dependence on export sales and a relatively elastic export demand (wd

is small and Ze is large), little is to be gained by advertising domestic sales.The supply elasticity directly a¡ects the optimal amount of advertising

only for the lump-sum funding model. Regardless of the funding arrange-ments, a more-elastic supply function implies a smaller price increase andsmaller bene¢ts to producers from a given advertising-induced demand shift.With levy funding, however, an increase in the supply elasticity implieschanges in the producer costs of a given advertising expenditure along withthe changes in the producer bene¢ts. The supply elasticity drops out of theadvertising intensity formulae because, at the pro¢t-maximising margin, theadvertising-induced price increase just o¡sets the increase in marginal costfrom the higher levy associated with the increased advertising. Moregenerally, for inframarginal advertising levels, the less elastic is supply thegreater will be the net price increase and increase in producer quasi-rentsresulting from the levy and advertising.An interesting result in table 1 is that there is essentially no di¡erence

between an ad valorem levy and a per unit levy. At the pro¢t-maximisinglevel of advertising, an ad valorem levy, t, is readily transformed to anequivalent per unit levy, T , and vice versa, using t � T =P and T � tP, whereP is price. Hence, the elasticity rules for the optimum advertising intensityare equivalent for the two types of levy.11 However, this similarity is a long-run or average outcome. Shifts in demand and supply curves from year to

11 This result, in particular, rests on the maintained assumption of competition. If, forinstance, processors were oligopsonistic, there might be signi¢cant di¡erences in impli-cations of per unit versus ad valorem levies for the distribution of the costs of the levy andthus for the producers' optimum. Zhang and Sexton (2000) have compared lump-sum andper unit levies to fund advertising with imperfectly competitive middlemen.



year, for example, because of seasonal conditions and the stage of thebusiness cycle, may mean that constant values for t and T over time wouldmean di¡erent patterns of annual advertising expenditures ö the particularvalue for t that is equivalent to a given value of T will vary from year to yearbecause of variations of price, P, and may be hard to determine ex ante.12

There are fundamental di¡erences between the producer pro¢t-maximisingadvertising expenditure when comparing lump-sum funding and levyfunding. Interestingly, more advertising is pro¢table with levy funding.13

This arises because both the statutory and ¢nal costs of lump-sum fundingare borne entirely by producers, whereas levy funding shifts the marginalcost and supply curve upwards, and as a result some of the levy costs arepassed on to buyers. The more elastic is product supply or product demand,the smaller is the price-increasing e¡ect of lump-sum funded advertising,and hence the smaller is the pro¢t-maximising advertising expenditure.A comparison of the formulae in table 1 for a closed economy model

and for an export model provides salutary warnings about applying theclosed economy model to derive pro¢t-maximising advertising and levy rateswhen the commodity of interest is traded. The formula for the levy rate onproduction to fund advertising of domestic sales of a good that is alsoexported is still equal to the overall demand elasticity with respect toadvertising divided by the overall elasticity of demand with respect to price.But these elasticities depend on market shares and price elasticities on thedi¡erent markets. To just apply elasticities for the domestic market wouldlead to an over-investment in domestic advertising, and the more so the moreimportant the export market and the more elastic export demand.

5. Policy interventions

In many parts of the world, and still in some industries in Australia,agriculture is subject to government policy interventions a¡ecting farm

12 In addition, if industry gross revenues are more stable than aggregate quantities(because of downward-sloping demand and supply variability being the main source ofvariability in prices and quantities), an ad valorem levy is likely to generate a more stablestream of revenues to fund promotion expenditures. Also, a constant per unit levy impliesdi¡erent percentage tax rates applying to di¡erent qualities of the same good at a point intime, which will distort the quality mix and implies some inequities. In spite of theseapparent advantages of ad valorem levies, per unit levies are widely used. One possibleexplanation is that per unit taxes might involve lower collection costs or less potential fortax evasion; obligations under an ad valorem tax can be reduced by under-reporting thevalue, for instance.

13 This point is noted by Goddard and McCutcheon (1993) and explained more clearlyin the elaboration by Kinnucan (1999a).



product prices (see, for example, Godden 1997). Interventions for tradedgoods can take the form of import restrictions via tari¡s, quotas andphytosanitary regulations, export subsidies, and price discrimination schemeswith pooling of returns to producers across markets or product type. Thesepolicy interventions are likely to a¡ect the optimal advertising strategy.Hence, we require a more elaborate model than those discussed above.Although the general structure, ¢rst-order conditions, and solution forms areessentially the same, regardless of the market situation, di¡erent marketstructures and policies and funding mechanisms imply di¡erent marketclearing conditions and thus imply di¡erent speci¢c solutions for optimaladvertising. In every case, however, as for the simpler models, producerbene¢ts from advertising arise only if the advertising raises the producerprice or average revenue.In many industries that engage in generic advertising programs, for

instance, the £uid milk industries in individual states in Australia and theUnited States, the relevant commodity market model would be the small-country trade model, whether we were thinking in terms of trade with othercountries or with other states within the same country. That is, the state ofVictoria ought to be regarded as a price taker in the national and worldmarkets for milk. As our results above show, in a setting of competitivemarkets and no policy interventions, advertising fresh milk could not bepro¢table for Victoria. The possibility of privately pro¢table product adver-tising is created through trade barriers that separate the Victorian £uid milkmarket from the £uid milk markets in other states, and the manufacturingmilk market. Recent changes, that deregulated the milk markets in Australia,might well have eliminated any such possibility. At a minimum these changesimply that serious scrutiny should be given to the question of further genericadvertising of milk and dairy products in Australia.In cases like this, when the potential for pro¢table advertising is entirely

a consequence of price policies that separate markets, the bene¢ts fromadvertising and the producer optimum depend on how the impact ofadvertising-induced demand growth is apportioned between price andquantity changes, which depends on the details of the policy (e.g., see Alstonet al. 1994). To illustrate, take the case of home consumption price schemesas originally modelled for dairy products by Parish (1962), and since appliedto many other products (e.g., see Alston and Freebairn 1988). Figure 4describes the simplest case. A relatively high domestic price Pd is set forhome sales, with Pd greater than the world price Pw. For simplicity assume asmall country so that Pw is ¢xed, and further that Pd is predetermined.Producers receive an average or pool price as shown by the pool priceline, Pp � Pw � Pd�Qÿ Qd�=Q, where Qd is domestic sales and Q is totalproduction.



Now, in ¢gure 4, suppose an increase in lump-sum funded advertisingshifts the domestic demand curve out from Dd

0 to Dd1. This raises the pool

price curve from Pp

0 to Pp

1 and the producer price and quantity outcome isgiven by point d rather than point c. Producers receive additional grosspro¢ts equal to area abdc, and optimal advertising would be at the pointwhere QdPp=dA � 1; that is, equation (7) but with the pool price Pp replacingthe competitive market price P.In short, using more elaborate and realistic models of the determination

of market prices and quantities to re£ect government policy interventions,the procedures of sections 2 and 3 can be extended readily to take intoaccount the e¡ects of various government policy interventions. Also, levyfunding of advertising and government subsidy funding can be added to themodel. Similar extensions can be used to incorporate elements of marketpower behaviour, for example as modelled in Goddard and McCutcheon(1993), Suzuki et al. (1994), and Zhang and Sexton (2000).

6. Multiple commodity effects

In some situations, the single-commodity, partial equilibrium model ofthe preceding sections might miss important cross-commodity price andadvertising e¡ects and provide misleading estimates of the pro¢t-maximisingadvertising level and levy rate. For example, advertising of one of the meats(or fruits) may directly shift the demand curve for the other meats (or fruits).Or, indirectly, the advertising-induced changes in prices of one meat (orfruit) is likely to have cross-commodity e¡ects on the demand for and supplyof other meats (fruits), and a set of price feedback e¡ects following thesechanges in turn have second-round e¡ects on market outcomes. These direct

Figure 4 Lump-sum funded advertising with a home consumption price scheme



and indirect second-round e¡ects for interrelated commodity prices will alterthe realised pro¢t for the advertised meat (or fruit), and for the other meats(fruits).Cross-commodity price and advertising e¡ects can be captured in a

multiple product model, for example, for the meats, fruits, dairy products,and so forth. Structural demand equations would include cross-commodityprices and advertising variables as well as the own-price and advertisingvariables. Structural supply equations similarly would include cross-commodity prices. Piggott et al. (1995) provide a good example of the formalmodelling of cross-commodity price and advertising e¡ects to obtain totalderivatives of prices and quantities with respect to advertising. Kinnucan(1996) and Kinnucan and Miao (2000) have used formal multiple commoditymodels in considering the implications of cross-commodity e¡ects onmeasures of the bene¢ts of advertising a particular product.For the advertisers of a particular commodity for which there are cross-

commodity e¡ects, the second-round e¡ects may increase or decrease thereturns to advertising, and thus the optimal advertising expenditures. Boththe advertising-induced demand increase and the levy cost e¡ect give rise toa higher price for the advertised product, which in turn gives rise to anincrease in the demand for substitute products and in their prices. Then,indirectly, there will be a second-round increase in demand for theadvertised commodity; for complementary products the converse happens.On the other hand, in the likely case where advertising of one commoditydirectly reduces demand for substitute commodities and in turn their pricesfall, this line of causation will reduce the demand for the advertisedcommodity and reduce the returns to its advertising. The net outcome ofthese potential positive and negative direct and indirect second-rounde¡ects, and the magnitude of the full e¡ects relative to the ¢rst-round e¡ectcaptured by the single-commodity model, become an empirical issuedepending on own- and cross-commodity elasticities of demand and supplywith respect to advertising and prices. Where advertising one commodityhas important direct and indirect cross-commodity e¡ects, the returns toother commodity producers are also altered, but the direction of e¡ect isambiguous.Discussion of the choice of pro¢t-maximising levels of advertising and

associated levy rates where there are important cross-commodity advertisingand price e¡ects raises a whole new set of issues related to strategic relationsby the di¡erent commodity decision makers. Here the returns to advertisingby one set of commodity producers depend on the advertising strategyadopted by advertisers of other commodities. In essence, we enter the worldof strategic games between duopolists and oligopolists with control overadvertising, but with competitive or price-taking supply response behaviour.



The choice of pro¢t-maximising strategies may be considered as varioustypes of cooperative and noncooperative games. Alston et al. (2000) providesome initial results along this line of analysis.

7. Some implications for econometric model specification

Use of the formulae summarised in table 1 to derive producer pro¢t-maximising amounts of generic advertising imposes a number of restrictionsthat might be treated as testable hypotheses in econometric models of thedemand response to advertising. In some senses the restrictions are morebinding than those required to make positive assessments of the e¡ects ofadvertising, and many published econometric studies have not allowed for ortested for some of the more subtle e¡ects of advertising. However, theseadditional restrictions on parameters or elasticities to be used to derivepro¢t-maximising levels of advertising add to the demands on data neededfor estimation.The derivation of pro¢t-maximising advertising expenditures requires

estimates of a demand function that is consistent with diminishing marginalreturns to advertising. This means that the demand function, Qd � f �A; :::�,must be at least twice di¡erentiable in the advertising expenditure variable,and that the second derivative must be negative.14 Estimates of simple linearfunctions of the form Qd � a0 � a1A, found in many studies do not meet thisrequirement. Such functions can be used to test hypotheses that advertisinga¡ects sales, and to assess whether current advertising levels are pro¢table ornot, but they are not useful in determining the pro¢t-maximising advertisinglevel. Richer, but nevertheless parsimonious, functional forms that allow adiminishing marginal sales response to advertising include Q � a0 ÿ a1=A,Q � a0 ÿ a1Q=A, and Q � a0 � a1 ln A, each of which is discussed and used inestimation by Goddard and McCutcheon (1993). Several studies have usedvariants of the square-root model: Q � a0 � a1A

1=2. The logarithmic functionln Q � a0 � a1 ln A allows for a diminishing marginal return to advertising,but it imposes a constant advertising elasticity. Alternatively, quadratic (andeven higher-order) advertising terms can be added to a linear function, suchas Q � a0 � a1Aÿ a2A

2. Compared with the simpler models, the extra termsrequire additional variation of advertising levels in the data if reliableestimates are to be made.In some cases, the transmission process from more advertising to higher

pro¢ts is via further di¡erentiation of the commodity and, in turn, a less

14General advertising response models sometimes posit zones of increasing returns toadvertising in addition to a zone of decreasing returns. Pro¢t maximisation requires that theadvertising expenditure be chosen in the zone of diminishing returns.



price elastic demand (Quilkey 1986 also discusses cases where the oppositee¡ect is achieved ö where the purpose of advertising is to position thecommodity more closely to a competitor, to increase the elasticity of demandö and discusses conditions when each strategy might be better). The shiftfrom D0 to D2 in ¢gure 1 provides an illustration. To test for this e¡ect ofadvertising requires the inclusion of a cross-product term for price andadvertising (for example, not just Q � aÿ bP� . . ., but also as Q �aÿ bP� cPA� . . .). Brester and Schroeder (1995), and several referencestherein, include some speci¢cations of this nature and they ¢nd limitedempirical support for the idea that advertising reduces the price elasticity ofdemand for some agricultural commodities. The addition of cross-price andadvertising explanatory variables adds to the required independent variationof price and advertising in the data if signi¢cant e¡ects of advertising onprice elasticities are to be detected. In a world in which it is challenging toobtain good and robust estimates of the direct e¡ects of advertising on theposition of the demand curve, it might be asking too much to try to measurealso the e¡ects on the slope of the curve, and the rates of change of theseresponses with respect to the total advertising budget.Among the econometric problems to be considered, issues of endogeneity

and simultaneity might be important when advertising is funded using levyfunds rather than lump-sum funding. Carman and Green (1993) discuss theissue of `supply response to promotion'. In our derivations above, there is astrict link between the levy-induced supply shifts and correspondingadvertising-induced demand shifts. In reality the statistical linkages may beweakened by the di¡erences in timing between collection of funds andexpenditure or actual advertising, and the fact that some levy funds arespent on other activities. It is su¤cient for now to note the potential forsuch problems, and the fact that the funding method might play a role indetermining their importance.The optimisation models might also be extended to allow for dynamic

e¡ects (for example, Nerlove and Arrow 1962).15 Most of the recent econo-metric studies of demand response to generic advertising have allowed forpersistence e¡ects, especially those studies using monthly or quarterly data.It is common to see ¢ndings where advertising e¡ects persist for severalperiods ö for instance, several quarters beyond the quarter in which theadvertising took place (or, at least, when the expenditure was made), evenfor products like milk or meat for which intrinsic demand dynamics are not

15 The models discussed in the article use single-period e¡ects of advertising. They readilycan be extended to include multiperiod e¡ects. Essentially, the partial advertising deriva-tives, or advertising elasticities, would be represented by the discounted or present value ofthe stream of current and future e¡ects of advertising.



usually important compared with other goods that have aspects of fashion(such as wine), addiction (such as tobacco), or durability (such as woollenapparel or cars). Estimates of the multiperiod e¡ects of advertising oncommodity demand require time-series or panel data sets for models withcurrent and lagged advertising variables as explanatory variables. Again, theextra explanatory variables, even when supported by restrictions on lagstructures, add to the required independent variation of advertising outlays.Of course, estimation of the e¡ects of advertising involves the usual long

list of speci¢cation, estimation and evaluation challenges. Attention needs tobe given to such factors as (1) allowing for the e¡ects of other explanatoryvariables such as prices, incomes, and non-advertising demand-shift vari-ables; (2) measurement of advertising; and (3) potential spurious correlationsfrom non-stationary dependent and explanatory variables. The volume byKinnucan et al. (1992) provides a review of some of the challenges and ofprogress so far.

8. Conclusion

We have collated rules for the generic advertising expenditures that willmaximise producer quasi-rents or pro¢ts in a competitive commodity marketsetting in which individual producers choose output quantities to equate theirmarginal cost with the market price. A single-market, single-period modelcan be generalised easily to allow for international trade, government policyinterventions, multiple commodity interactions, and to incorporate dynamicand lagged responses of sales to advertising.E¡ective advertising shifts out the commodity demand curve and also

may make demand less (or more) elastic. In order to raise producer pro¢ts,advertising must cause the producer price to increase. In many instances thisoutcome would not be possible without government intervention, theimposition of price policies that allow markets to be separated and preventarbitrage from eliminating any advertising-induced price increases. This issurely the case for state-level generic milk advertising programs, which are inaggregate perhaps the most economically important generic commodityadvertising programs. The pro¢t-maximising advertising expenditure equatesmarginal returns, given by the increase in average producer revenue (simplyprice in the case of undistorted markets) times output, with the marginal costto producers of the advertising.Four di¡erent methods of funding generic commodity advertising were

considered. Much of the literature focuses on lump-sum funding, however,this seems more applicable to industries with market power than the com-petitive industry structure considered in this article. Levies on producersmade compulsory by supporting government legislation are more common



in practice. A per unit or ¢xed levy and an ad valorem levy were found tobe equivalent at the pro¢t-maximising level of advertising. The levye¡ectively shifts the supply curve up, and a portion of the incidence is passedon to buyers as higher prices; in contrast, under the lump-sum fundingoption, costs of advertising are fully borne by producers. A fourth fundingoption allows for a government subsidy or matching grant for producerfunds collected for advertising.Generally, the e¡ects of the subsidy are to raise the pro¢t-maximising total

advertising budget and to lower the producer levy rate. In other words,relative to the lump-sum funding option, the use of levies shifts some of themarginal cost of advertising onto consumers, and the use of subsidies shiftssome of the marginal cost onto taxpayers, and both of these elementsmake the advertising more pro¢table for producers; hence, their optimaladvertising expenditure is higher.In our analysis of levy-funded advertising, we imposed a restriction that

the levy funds were fully spent on advertising in the same period in which thefunds were collected. This is a very strong link imposed between the fund-raising and expenditure sides of the problem, which might not be matchedexactly in practice, even in the absence of matching government grants. Inpractice, funds must be spent after they are raised (unless some kind of debt¢nance is being used), some funds are spent in administration or on activitiesother than advertising (including production research, market research,market information, public relations, other forms of promotion), and somefunds are carried forward into future periods. This means that the linkage inpractice between taxing and spending is not as tight as we have formallymodelled, and the real-world situation might fall in some senses in betweenlump-sum and levy funding as we have modelled them.The advertising intensity, the advertising budget as a share of gross sales,

equals the ad valorem levy rate. At the pro¢t-maximising level of advertisingfunded by a levy, a Dorfman^Steiner rule is found: the advertising intensityand the levy rate are equal to the elasticity of quantity demanded withrespect to advertising divided by the (absolute value of the) price elasticity ofdemand. Then, when the demand for a commodity is more responsive toadvertising, or less responsive to price, the price increase and bene¢ts toproducers from more advertising will be greater, and therefore so will theoptimal levy and the advertising budget be greater.Application of the pro¢t-maximising advertising rules requires quantitative

information on the demand response to advertising and price, not just theaverage impact multipliers or elasticities, but information on how those multi-pliers and elasticities vary with the quantity of advertising. Diminishingmarginal returns to advertising is a necessary condition for de¢ning a maxi-mum, and meeting this requirement places extra demands on the speci¢cation,



data, and estimation of the demand function. In particular, the demandequation needs to be twice di¡erentiable in the advertising explanatory variable,and interaction terms between price and advertising are required if we wish totest for advertising-induced changes in the price elasticity or slope of demand.Our results have some direct and indirect policy implications. We have

observed that generic advertising cannot pay in the small open-economy casein the absence of government intervention in the market. This observationraises questions about the future pro¢tablity, in a deregulated market, of thetypes of generic milk promotion programs that have been important inAustralia to date. The US government has recently introduced a requirementfor regular assessment of the consequences of commodity promotion schemesfunded by mandatory levies, under federal marketing orders. Although wedo not yet have such requirements in Australia, it would seem reasonable tointroduce some conditions when mandatory levies are being collected to fundgeneric advertising, sanctioned by the government. For instance, at aminimum, we could require the organisation in question to collect data andmake it available, to permit an independent assessment of the e¡ects of theadvertising and the levy. This would improve the information £ow toproducers and to policy makers for future decision-making.

References

Ackerman, K.Z. and Henneberry, S.R. 1992, Èconomic impacts of export marketpromotion', in Armbruster, W.J. and Lenz, J.E. (eds), Commodity Promotion Policy in aGlobal Economy, Proceedings of the NEC-63 Fall Symposium, Arlington, Virginia, 22^3October 1991, pp. 46^59.

Alston, J.M., Carman, H.F. and Chalfant, J.A. 1994, Èvaluating primary productpromotion: the returns to generic advertising by a producer cooperative in a small, openeconomy', in Goddard, E.W. and Taylor, D.S. (eds), Promotion in the Marketing Mix:What Works and Why?, Proceedings from the NEC-63 Spring '94 Conference in Toronto,Ontario, 28^9 April.

Alston, J.M., Carman, H.F., Chalfant, J.A., Crespi, J., Sexton, R.J. and Venner, R.S.1998, The California Prune Board's Promotion Program: An Evaluation, GianniniFoundation Research Report No. 344, University of California, Davis, California.

Alston, J.M. and Freebairn, J.W. 1988, `Producer price equalization', Review of Marketingand Agricultural Economics, vol. 56, pp. 306^39.

Alston, J.M., Freebairn, J.W. and James, J.S. 2000, `Beggar thy neighbour advertising:theory and application to generic commodity promotion programs', mimeo, Universityof California, Davis.

Brester, G. and Schroeder, T. 1995, `The impacts of brand and generic advertising on meatdemand', American Journal of Agricultural Economics, vol. 77, pp. 969^79.

Carman, H.F. and Green, R.D. 1993, `Commodity supply response to a producer-¢nancedadvertising program', Agribusiness, vol. 9, pp. 605^21.

Chang, H.-S. and Kinnucan, H.W. 1991, Èconomic e¡ects of an advertising excise tax',Agribusiness, vol. 7, pp. 165^73.



Cran¢eld, J. and Goddard, E. 1999, Òpen economy and processor oligopoly power e¡ectsof beef advertising in Canada', Canadian Journal of Agricultural Economics, vol. 47,pp. 1^9.

De Boer, J. 1997, `Rural product promotion: economic aspects of promotability,organisation and public assistance', Review of Marketing and Agricultural Economics,vol. 45, pp. 121^45.

Dorfman, R. and Steiner, P. 1954, Òptimal advertising and optimal quality', AmericanEconomic Review, vol. 44, pp. 826^36.

Forker, O. and Ward, R. 1993, Commodity Advertising: The Economics and Measurementof Generic Programs, Lexington, New York.

Goddard, E.W. and Conboy, P. 1993, Òptimal international promotion expenditure fordi¡erentiated products', Review of Marketing and Agricultural Economics, vol. 61,pp. 49^61.

Goddard, E.W. and McCutcheon, M. 1993, Òptimal producer investment in genericadvertising: the case of £uid milk in Ontario and Quebec', Canadian Journal ofAgricultural Economics, vol. 41, pp. 329^47.

Goddard, E.W. and Tielu, A. 1988, Àssessing the e¡ectiveness of milk advertising inOntario', Canadian Journal of Agricultural Economics, vol. 36, pp. 261^78.

Godden, D. 1997, Agricultural and Resource Policy: Principles and Practice, OxfordUniversity Press, Melbourne.

Hayes, D.J. and Jensen, H.H. 1992, `Generic advertising without supply control:implications of mandatory assessments', in Armbruster, W.J. and Lenz, J.E. (eds),Commodity Promotion Policy in a Global Economy, proceedings of the NEC-63 FallSymposium, Arlington, Virginia, 22^3 October 1991, pp. 90^104.

Hill, D., Piggott, R. and Gri¤th, G. 1996, `Pro¢tability of incremental expenditure on ¢brepromotion', Australian Journal of Agricultural Economics, vol. 40, pp. 151^74.

Industries Assistance Commission 1976, Financing Promotion of Rural Products, AGPS,Canberra.

Kinnucan, H.W. 1996, À note on measuring returns to generic advertising in interrelatedmarkets', Journal of Agricultural Economics, vol. 47, pp. 261^7.

Kinnucan, H.W. 1997, `Middleman behaviour and generic advertising rents in competitiveinterrelated industries', Australian Journal of Agricultural and Resource Economics,vol. 41, pp. 191^207.

Kinnucan, H.W. 1999a, Àdvertising traded goods', Journal of Agricultural and ResourceEconomics, vol. 24, pp. 38^56.

Kinnucan, H.W. 1999b, Òptimal generic advertising decisions in supply-managedindustries: clari¢cation and some further results', Canadian Journal of AgriculturalEconomics, vol. 47, pp. 157^66.

Kinnucan, H.W., Chang, H.-S. and Venkateswaran, M. 1993, `Generic advertising wearout',Review of Marketing and Agricultural Economics, vol. 61, pp. 401^15.

Kinnucan, H.W. and Christian, J.E. 1997, À method for measuring returns to nonpriceexport promotion with application to almonds', Journal of Agricultural and ResourceEconomics, vol. 22, pp. 120^32.

Kinnucan, H.W. and Miao, Y. 2000, `Distributional impacts of generic advertising onrelated commodity markets', American Journal of Agricultural Economics, vol. 82,pp. 672^8.

Kinnucan, H.W. and Myrland, Ò. 2000, Òptimal advertising levies with application to theNorway-EU Salmon Agreement', European Review of Agricultural Economics, vol. 27,pp. 39^57.



Kinnucan, H.W., Thompson, S. and Chang, H.-S. (eds) 1992, Commodity Advertising andPromotion, Iowa State University Press, Ames, Iowa.

Lambin, J.J. 1976, Advertising, Competition and Market Conduct in Oligopoly over Time,Amsterdam, North-Holland.

Nerlove, M. and Arrow, K. 1962, Òptimal advertising policy under dynamic conditions',Economica, vol. 29, pp. 129^47.

Nerlove, M. and Waugh, F. 1961, Àdvertising without supply control: some implicationsof a study of the advertising of oranges', Journal of Farm Economics, vol. 43,pp. 813^37.

Nichols, J.P., Kinnucan, H.W. and Ackerman, K.Z. (eds) 1991, Economic E¡ects of GenericPromotion Programs for Agricultural Exports, proceedings of the NEC-63 SpringSymposium, Washington, DC, 22^3 February 1990.

Parish, R. 1962, `The costs of protecting the dairy industry', Economic Record, vol. 82,pp. 167^82.

Piggott, R., Piggott, N. and Wright, V. 1995, Àpproximating returns to incrementaladvertising expenditures: methods and applications to the Australian meat industry',American Journal of Agricultural Economics, vol. 77, pp. 497^511.

Quilkey, J. 1986, `Promotion of primary products ö a view from the cloister', AustralianJournal of Agricultural Economics, vol. 30, pp. 38^52.

Suzuki, N., Kaiser, H.M., Lenz, J.E., Kobayashi, K. and Forker, O.D. 1994, Èvaluatinggeneric milk promotion e¡ectiveness with an imperfect competition model', AmericanJournal of Agricultural Economics, vol. 76, pp. 296^302.

Vande Kamp, P. and Kaiser, H. 1999, Commodity Promotion Programs in the United States,Department of Agricultural, Resource and Managerial Economics, Cornell University,Ithaca, NY, May.

Wohlgenant, M. 1993, `Distribution of gains from research and promotion in multistageproduction systems: the case of US beef and pork industries', American Journal ofAgricultural Economics, vol. 75, pp. 642 ^51.

Zhang, M. and Sexton, R.J. 2000, Òptimal commodity promotion in imperfectlycompetitive markets', mimeo, Department of Agricultural and Resource Economics,University of California, Davis.



generic advertising without supply control: implications of funding

Documents